From owner-freebsd-arm@FreeBSD.ORG Fri Mar 13 12:35:02 2015 Return-Path: Delivered-To: freebsd-arm@freebsd.org Received: from mx1.freebsd.org (mx1.freebsd.org [8.8.178.115]) (using TLSv1.2 with cipher AECDH-AES256-SHA (256/256 bits)) (No client certificate requested) by hub.freebsd.org (Postfix) with ESMTPS id 6CBC6865 for ; Fri, 13 Mar 2015 12:35:02 +0000 (UTC) Received: from gromit.dlib.vt.edu (gromit.dlib.vt.edu [128.173.126.120]) (using TLSv1.2 with cipher ECDHE-RSA-AES256-GCM-SHA384 (256/256 bits)) (Client CN "gromit.dlib.vt.edu", Issuer "Chumby Certificate Authority" (not verified)) by mx1.freebsd.org (Postfix) with ESMTPS id 41EB8F7 for ; Fri, 13 Mar 2015 12:35:02 +0000 (UTC) Received: from pmather.lib.vt.edu (pmather.lib.vt.edu [128.173.126.193]) (using TLSv1 with cipher ECDHE-RSA-AES256-SHA (256/256 bits)) (No client certificate requested) by gromit.dlib.vt.edu (Postfix) with ESMTPSA id 1DE57416; Fri, 13 Mar 2015 08:35:00 -0400 (EDT) Content-Type: text/plain; charset=us-ascii Mime-Version: 1.0 (Mac OS X Mail 7.3 \(1878.6\)) Subject: Re: BeagleBone slow inbound net I/O From: Paul Mather In-Reply-To: <20150312232641.4365263d@ivory.wynn.com> Date: Fri, 13 Mar 2015 08:34:59 -0400 Content-Transfer-Encoding: quoted-printable Message-Id: References: <20150311165115.32327c5a@ivory.wynn.com> <20150312133433.GB28385@cicely7.cicely.de> <20150312232641.4365263d@ivory.wynn.com> To: Brett Wynkoop X-Mailer: Apple Mail (2.1878.6) Cc: "freebsd-arm@freebsd.org" , Bernd Walter X-BeenThere: freebsd-arm@freebsd.org X-Mailman-Version: 2.1.18-1 Precedence: list List-Id: "Porting FreeBSD to ARM processors." List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , X-List-Received-Date: Fri, 13 Mar 2015 12:35:02 -0000 On Mar 12, 2015, at 11:26 PM, Brett Wynkoop = wrote: > Greeting- >=20 > It is confirmed, the net I/O is much slower than reading from the SD > card. Here is another run of tar - pipe - tar, but this time the > source is the sd card and the destination is still the USB zfs. The Beaglebone uses a USB NIC so any I/O to/from SD card will be = competing for resources and potentially slowing things down. Have you tried outputting to /dev/null to get a more accurate sense of = how performant the network I/O alone is? Alternatively, how about using = something like benchmarks/netperf? I just tried transferring a big file from another system to my = Beaglebone Black via "nc", piping the output to /dev/null on the = Beaglebone Black, and I am maxing out the 100 Mbit connection: pmather@beaglebone:~ % netstat -w 1 -I cpsw0 input cpsw0 output packets errs idrops bytes packets errs bytes colls 0 0 0 12325666 0 0 269572 0 0 0 0 12302830 0 0 268372 0 0 0 0 12343768 0 0 291730 0 0 0 0 12281634 0 0 267910 0 0 0 0 12284722 0 0 267976 0 0 0 0 12320998 0 0 268768 0 0 0 0 12299862 0 0 268306 0 0 0 0 12290778 0 0 268150 0 0 0 0 12315002 0 0 268636 0 0 0 0 12269522 0 0 267646 0 0 0 0 12322572 0 0 268834 0 0 0 0 12328568 0 0 268900 0 0 0 0 12311974 0 0 268570 0 0 0 0 12281634 0 0 267910 0 0 0 0 12302890 0 0 268372 0 0 0 0 12304344 0 0 268438 0 0 0 0 12289264 0 0 268042 0 0 0 0 12311914 0 0 268570 0 0 0 0 12302890 0 0 268372 0 0 0 0 11554466 0 0 252070 0 0 0 0 12315002 0 0 268636 0 input cpsw0 output packets errs idrops bytes packets errs bytes colls 0 0 0 12307372 0 0 268616 0 0 0 0 12289264 0 0 268042 0 0 0 0 12314942 0 0 268636 0 0 0 0 12299862 0 0 268306 0 ^C So, from my experience, inbound network I/O is as good as can be = expected, at least on the Beaglebone Black. Cheers, Paul.=